8 practices · updated 2026-08-23
Field notes.
What I have found running buildings, and what I built to run them better. Each practice is here because something forced it — a roster that would not reach full time, a department that turned out to be mostly trainees, a number I published and had to withdraw.
Most of it is not really about hotels. A rota that cannot reach full time is a scheduling problem; stock that is counted twice and trusted once is an inventory problem; a department that is two-thirds trainees is an HR problem. They turn up the same way on a site, in a warehouse, or anywhere a small team runs something physical.
The practices are here. The figures behind them are not: they belong to the properties and the people who work in them.
Hours and retention
01 · The single biggest retention lever a small operator controls
Nobody leaves over pay. They leave over hours.
A small property cannot give a housekeeper forty hours. There is not enough work in the building. So the job is structurally part-time, and a part-time job loses to any full-time offer in town — even a worse-paid one.
This is arithmetic, not a management failure, and it cannot be fixed inside one property. It can be fixed across two: properties within driving distance can build one full-time schedule out of two part-time ones.
02 · Coverage · single point of failure
One person over forty hours is a risk, not a saving.
When one supervisor absorbs the overtime, the property runs on that person's continued availability. No cover for illness, none for leave, and no second person who knows the close.
Track hours per person, not just total hours. A department at the right headcount can still be one absence from failing.
03 · Advancement
The ladder is the retention mechanism, and it only exists at group scale.
A twelve-person property has exactly one rung to climb. When it is occupied, everyone below has nowhere to go, and the ambitious leave first — the opposite of who you want leaving.
Across a group that rung exists once per property. Post internal openings to every property before advertising outside. It costs nothing, and it is the only promise about the future a small operator can actually make.
04 · Report it separately or it disappears
Losing the year-round core is expensive in a way the roster does not show.
Seasonal churn is normal and planned for. Losing the people who stay all year is different: they hold the knowledge that is written down nowhere — which supplier to call, which room floods, how the close actually works.
Report core turnover separately from seasonal turnover. Blended into one number it vanishes, because seasonal churn is large and expected.
How to manage by numbers
07 · Their number is not arguable. An imported one is.
Use the property's own standard, never an industry one.
Every operation already prints its planning constants somewhere — on the schedule board, the assignment sheet, the posted notice. Find it before importing a benchmark.
A manager can dispute an industry average. They cannot dispute the number in their own handwriting.
08 · Self-evidencing measurement
Two numbers in the same column beat either one alone.
"Attendants needed: 17" is a plan. "Attendants scheduled: 11" is a fact. Side by side, written by the same person on the same day, they leave nothing to dispute and need no interpretation.
Design every report so the plan and the actual sit next to each other. A report showing only actuals invites an argument about what the target was.
09 · An error we made, published, and withdrew
Count what is printed, never what is handwritten.
Boards get photocopied and marked up. Handwritten names usually mean who checked this, not who worked this. Counting them as staff distorts everything downstream.
Before quoting any staffing ratio, confirm which rows were printed. One counting error propagates into every figure derived from it.